Distributed Map Integration For Multi-robot Exploration

Building map after exploring unknown environment is a classic problem in mobile robot research. From the task performer's point of view, research hotspots in recent years have moved from single-task robots to work together in multi-robots approach. Multi-robots divide an unknown area into several sub-regions to explore parallel. This way can improve the efficiency of the task execution but bring new challenge– map integration. Most of actual systems are short of effective location measures, so make sub-region map information difficult to be merged into an accurate and complete map. Map integration can solve the map transformation and matching. Therefore, the study of map integration technique is important for multi-robots collaborative exploring environment.The topic comes from National Natural Science Foundation program" The network environment based on swarm intelligence approach virtual machine Collaborative Platform" The program hopes to explore a new parallel computing model to control the large-scale group collaboration and verify performance through map building classic scene. Scenarios of the program hope robots randomly scattered in the environment, and the hardware does not have global positioning capabilities. In order to solve integration problems under such conditions, the topic optimized a line-segment based map fusion algorithm.The paper's main achievements include:1. Line-segment based map fusion algorithm: Use line-segments as the basic unit of information describing obstructions and extract the line geometric relationship as reference in the intersection region. Through lots of map transformation, two partial maps can be combined into the same coordinate system. Research adding spatial characteristics– relative distance between the obstacles can enrich the feature information, reduce the frequency of map transformation and improve efficiency of algorithm.2. Distributed map integration strategy: The paper uses diffusion method to achieve the process of distributed map fusion in network. Robots can be out of central control, merge maps and unify coordinate by themselves. On this basis, research introduces the credibility to improve the integration accuracy.3. Algorithm validation simulation: Research design and develop simulation in order to verify algorithm. This simulation emulates exploration without localization and map integration process. By testing integration effect in a variety of structured terrains, paper proved the algorithm can solve map integration under no localization. Then the data comparison of the same terrain reflects the optimized algorithm can improve the implementation efficiency of the integration. Finally, distributed map integration strategy achieves self-integration within robot group and benefits large scale multi-robots exploration application.